Re: Re: PHP 4.0 Bug #3647 Updated: Change to Global variablein Function Fails

Stanislav Malyshev wrote: > > AP>> You have to be kidding me!?!? This is still a bug. The variable was > > No, I'm not kidding you and this is not a bug. Declaring variable as > "global" makes local variable to be pointing to global one (like, $var = > &$GLOBALS["var"]). Unsetting this local one is unsetting this local one, > but the global one doesn't vanish. However, if you change it, global value > gets changed, because local one is pointing to global one. But, if you > assign to it a new reference, old one gets removed, so you again might not > get what you expect. I know, this is not easy to understand, but that's > the way references work in PHP now. Ahhh, thanks for the clarification. I did not know that a GLOBAL declaration is treated as a reference; I thought you would directly access the global variable. From reading 'language.variables.scope.html' one does not become aware of this distinction (i.e., that $GLOBALS["a"] !== GLOBAL $a). > > You can read Andi's article on reference counting at > http://www.zend.com/zend/art/ref-count.php > to get more understanding of this. Or read the source (not for the faint > of heart). I understand reference counting; I just was not aware that a GLOBAL declaration gives you a reference rather than the real thing. Obviously it was not that way in PHP3 as there were no reference assignments in PHP3 :-). > > AP>> declared as GLOBAL, so I shouldn't have to use the $GLOBALS array to set > AP>> *or* *unset* it. > > Well, you can try and sue PHP Group over violation your rights not to use > $GLOBALS, but this is how it presently works in PHP. To be fair, I'm not talking about suing, I filed a bug report :-). > If you don't like it, > submit feature/change request. I just trying to hel you to make it work > now. I appreciate that. Perhaps this can be changed into a documentation bug report ;-). OTOH, I think a lot of people will get bitten by this design implementation, so perhaps people will consider having a GLOBALy declared variable directly access the GLOBALS hash table. Thanks, Andreas